AutoGen
★ 59.4k · GitHub →Microsoft's pioneering multi-agent conversation framework — now in maintenance mode, succeeded by Microsoft Agent Framework.
Install
$ pip install autogen-agentchat What is AutoGen?
AutoGen was the framework that made "agents talking to each other" a mainstream pattern. Coming out of Microsoft Research, it modeled applications as conversations between agents — an AssistantAgent that writes code, a UserProxyAgent that executes it, group chats where a manager routes turns between specialists. The v0.4 rewrite rebuilt it on an async, event-driven actor architecture, split into layered packages (autogen-core, autogen-agentchat, autogen-ext), and added AutoGen Studio for no-code prototyping.
The critical fact for anyone evaluating it now: AutoGen is officially in maintenance mode. Microsoft has designated Microsoft Agent Framework (MAF) as its enterprise-ready successor — merging AutoGen with Semantic Kernel — and publishes a migration guide. AutoGen will receive no new features and is community-managed going forward. The repository remains popular and the CC-BY-4.0-licensed code still works, but the center of gravity has moved.
That history still matters. Years of research prototypes, papers, and tutorials were built on AutoGen, and its conversation-centric design remains one of the clearest mental models for multi-agent systems. But for a new production project in 2026, choosing AutoGen means choosing a framework whose own maintainer points you elsewhere — a hard thing to justify against actively developed alternatives.
Key capabilities
- ▸ Conversation-centric multi-agent model: assistants, user proxies, and managed group chats
- ▸ Async event-driven actor runtime (v0.4+) with layered core/agentchat/extension packages
- ▸ Built-in code execution agents that write, run, and debug code in sandboxed environments
- ▸ AutoGen Studio: a no-code GUI for prototyping and testing agent teams
- ▸ Large body of research, papers, and community examples accumulated since 2023
When to use it
Choose AutoGen only if you are maintaining an existing AutoGen codebase or reproducing research built on it — its conversation patterns and code-execution agents are still instructive, and the community keeps the lights on. Do not start a new production system on it: Microsoft directs new users to Microsoft Agent Framework, and actively developed alternatives (LangGraph for durable orchestration, CrewAI for role-based teams, OpenAI Agents SDK for lightweight handoffs) all have momentum AutoGen no longer has.
Alternatives
Frequently asked questions
Is AutoGen still maintained in 2026?
Only in maintenance mode. Microsoft announced AutoGen will receive no new features and is community-managed going forward. New users are officially directed to Microsoft Agent Framework, the successor that merges AutoGen with Semantic Kernel, and a migration guide exists for moving existing AutoGen code over. Bug-level upkeep continues, but active development has ended.
What replaced AutoGen?
Microsoft Agent Framework (MAF). It is Microsoft's production-ready agent framework, unifying AutoGen's multi-agent orchestration ideas with Semantic Kernel's enterprise features, with stable APIs, multi-provider model support, and A2A/MCP interoperability. Microsoft publishes an official AutoGen-to-MAF migration guide for existing projects.
What was AutoGen best known for?
Making multi-agent conversation a practical programming pattern. Its AssistantAgent/UserProxyAgent pairing — one agent writes code, another executes it and reports results — and its managed group chats influenced nearly every multi-agent framework that followed. It was especially popular in research settings for rapidly prototyping agent collaboration experiments.
Star count updated 2026-07-02 · refreshed weekly from the GitHub API · browse all tools →